---
title: What Macro Systems Can You Build with the Helious API and MCP Server?
description: Seven macro systems to build on the Helious API and MCP server, from a regime classifier to an AI morning brief, with code run on 17 November 2026.
source: Helious
canonical: "https://helious.io/blog/macro-systems-you-can-build-with-helious-in-2026"
---

# What Macro Systems Can You Build with the Helious API and MCP Server?

Helious · 18 November 2026 · 7 min read

Seven, at least: a regime classifier, a curve monitor, a risk gauge, a policy tracker, a supply monitor, an event-risk layer and a morning brief written by an AI agent. Each answers the same question on a schedule: which state are growth, inflation, policy and the rates market in, and has it changed? In 2026 there are two ways to build them on Helious. Write code against the REST API, or hand the same data to an agent as 27 MCP tools. This post runs one of the seven on the real numbers of 17 November 2026 and says what each costs.

## 1. What does a macro system need from its data?

A level is not enough. A system has to know whether 3.12% is rising or falling, and whether a VIX of 25 is high for this year. Helious returns the reading with the level:

- **Rates of change, worked out.** Every inflation measure carries its year rate and its 1, 3 and 6-month annualised rates. Every jobs measure carries 1, 3, 6 and 12-month windows.

- **The regime, named.** The Treasury curve arrives as bull or bear, steepener or flattener, with the time it flipped and every past flip.

- **A percentile.** Each volatility gauge carries its place in the past year. A release level carries its percentile across its own history.

- **A read that shows its working.** `/v1/context` names a cause for the session only when a scored release printed in the window. It lists the checks the read passed and what it does not cover.

## 2. Which macro systems can you build?

| System | The question it answers | Endpoints | MCP tools |
| --- | --- | --- | --- |
| Regime classifier | Is inflation heating, is labour slowing, is policy tightening? | `/v1/inflation`, `/v1/jobs`, `/v1/rate-odds`, `/v1/series/gdpnow` | get_inflation, get_jobs, get_rate_odds, get_series |
| Curve monitor | Which curve regime is the market in, and when did it flip? | `/v1/rates`, `/v1/curve` | get_rates_and_curve, get_curve_history |
| Risk gauge | Is risk calm or stretched, and are the usual links holding? | `/v1/vol`, `/v1/correlations`, `/v1/context` | get_vol_board, get_correlations, get_market_context |
| Policy tracker | What is priced for the Fed, and how does the committee lean? | `/v1/rate-odds`, `/v1/rate-path` | get_rate_odds, get_rate_path, get_fed_roster, get_document_digests |
| Supply monitor | How much is the Treasury borrowing, and how is it being bought? | `/v1/treasury-issuance`, `/v1/auctions` | get_treasury_issuance, get_auctions |
| Event-risk layer | Is a geopolitical or energy shock building? | `/v1/feed?topics=geo,energy`, `/v1/hormuz` | get_feed, get_hormuz_transits |
| Morning brief | What changed since yesterday, in plain English? | All of the above | get_build_plan, then the tools above |

The first six are layers of one system. Start with the classifier, then add a layer whenever the state needs one more word.

## 3. How do you build the regime classifier?

Four calls and three comparisons. All four calls answer with no key, so this runs as written. One caveat: with no key a new print is held for its first hour, so `growth` can read `None` just after an update.

```
import requests

API = "https://api.helious.io"

def data(path, **params):
    r = requests.get(API + path, params=params, timeout=15)
    r.raise_for_status()
    return r.json()["data"]

pce = data("/v1/inflation", measures="core_pce")["measures"][0]
jobs = data("/v1/jobs", measures="payrolls")["measures"][0]
odds = data("/v1/rate-odds")
now = data("/v1/series/gdpnow")["latest_card"]
priced = odds["path_total"]["change_bps"]

state = {
    "inflation": "heating" if pce["a3"] > pce["yoy"] else "cooling",
    "labour": "slowing" if jobs["m3"] < jobs["m12"] / 12 else "steady",
    "policy": "tightening" if priced > 0 else "easing",
    "growth": now["actual"],
}
print(state)
```

On 17 November 2026 it printed `{'inflation': 'cooling', 'labour': 'steady', 'policy': 'tightening', 'growth': '3.4%'}`. Read each label against the number behind it:

- **Cooling.** Core PCE ran at 2.90% annualised over the three months to September, under its 3.12% year rate.

- **Steady.** Payrolls averaged 137,000 a month over three months, against about 74,000 a month over twelve. The labour market was picking up, not slowing.

- **Tightening.** The fed funds strip priced 80.5 basis points of hikes through 15 September 2027.

- **Growth.** GDPNow had the fourth quarter tracking 3.4%.

One label is not a regime. On the same day core CPI said the opposite of core PCE: 3.75% annualised over the three months to October, against a year rate of 2.57% for September, the newest it has. A system worth running reads both and flags the disagreement. That costs one more measure in the same call: `measures=core_pce,core_cpi`. The answer lists measures in its own order, core CPI first, so pick each one by its `key`, not by position.

## 4. What does the rates market add to the state?

Four more calls, and the state covers the market as well as the economy. On 17 November they read:

- **Curve** (`/v1/rates`, `/v1/curve`). 2s10s stood at 30.3 basis points. Over five sessions the curve bear flattened: the 2-year rose 24 basis points and the 30-year rose 20. The last intraday flip was to a bull steepener at 15:32 UTC, out of a bear flattener. The [curve regimes are explained here](https://helious.io/blog/yield-curve-steepening-and-flattening-explained).

- **Volatility** (`/v1/vol`). VIX at 25.46 sat in the 94th percentile of the past year, and the board labelled it stressed. MOVE, the Treasury gauge, sat in the 93rd.

- **Session** (`/v1/context`). The read was event-adjacent: a release had landed, and the tape had not confirmed it as the driver. The daily moves of the 10-year yield and the S&P 500 had a correlation of -0.01 over 45 sessions, so neither a growth story nor a rates story was in charge.

These boards are market prices, so they follow the plan. They are live on Pro and an hour behind on a free key. With no key the boards answer a note in place of numbers.

## 5. Can an AI agent build it over the MCP server?

Yes, and the agent needs no code. Add the server once. In Claude Code that is one line:

```
claude mcp add --transport http helious https://api.helious.io/mcp
```

Helious is also listed in the [Claude connector directory](https://helious.io/blog/helious-in-the-claude-and-chatgpt-directories) and the [ChatGPT plugin directory](https://helious.io/blog/helious-is-now-a-plugin-in-chatgpt). Then give the agent the job in words: classify the US macro regime, say what changed since yesterday, quote each number and link its page. Three things make that work:

- **A plan before the first call.** `get_build_plan` takes the goal and a cadence. It returns the tools in order, the fields to read and the monthly cost, and it draws no credit. Asked on 17 November for a macro regime dashboard run hourly, it answered that one pass is 8 credits, about 5,760 a month.

- **A sentence to quote.** Every tool opens with one. That day `get_rate_odds` opened: "For the 2026-12-09 FOMC meeting the market prices 93% odds of a 25bp hike."

- **A page to cite.** Each answer names the Helious page the number lives on, so the brief links back to it.

Run that prompt on a timer and the seventh system, the morning brief, is built. The [AI setup post](https://helious.io/blog/how-to-connect-your-ai-to-helious-data) covers the other clients.

## 6. What does it cost to run?

A data call is 1 credit. One pass of the full recipe is 8 calls, so the cadence sets the bill:

- **Once a day:** about 240 credits a month.

- **Every hour:** about 5,760 credits a month.

Pro is $39.99 a month with 3,000 credits, every price live and 24 months of history. That covers about 12 full passes a day. Ultimate is $99 a month with no credit cap and 10 years of history, which is the plan for an hourly system. To wire it up first, a free account key has 100 credits a month, and with no key a network address shares 50. `get_build_plan`, `get_access_info` and `list_series` draw no credit on any plan.

Cache what does not move. The inflation and jobs boards change on a print, not on every call.

## 7. What are the limits?

- **The label is your rule.** Helious supplies the readings. What "cooling" means, and what a state favours, is the strategy you write and test.

- **Test on the consensus record, run the scores forward.** Helious holds each release's consensus, first actual, previous and revised prior as far back as your plan reads: 24 months on Pro, 10 years on Ultimate. Helious's own captured prints and scores start in mid-2026.

- **The depth is US.** The calendar also covers the euro area, the UK, Germany and Japan. Rates outside the US, options positioning and intraday credit spreads are not covered, and `/v1/context` says so in its answer.

- **A plan covers your own use.** Showing Helious data to your own users or customers is redistribution. It is in no plan and needs Ultimate with a separate written agreement.

## 8. Where do I start?

Run one call with no account: `curl "https://api.helious.io/v1/inflation?measures=core_pce"`. Then open the [macro system recipe](https://helious.io/developers/use-cases/#macro-system) for every call in order. Two more posts in this set cover the [trading algorithms](https://helious.io/blog/trading-algorithms-you-can-build-with-helious-in-2026) and the [macro forecasts](https://helious.io/blog/macro-forecasts-you-can-build-with-helious-in-2026) you can build on the same data.

The macro system recipe: every call in order, the fields to read and a working sample. The first calls need no key.

[See the macro system recipe →](https://helious.io/developers/use-cases/#macro-system)
